One of my favorite websites to use while teaching statistics is Seeing Theory, by Brown University. It is a wonderful website, because it allows people to “see” statistics.
Visualize concepts in statistics, to use the technically correct term, but you see what I mean.
One of the many reasons I like this website is because it presents a fun, interactive way to “get” what statistics is all about. It is one thing to talk about flipping a coin, it is quite another to actually flip a coin 1000 times. Or roll a dice, or understand what a probability distribution is, or to (finally!) “get” what the Central Limit Theorem is trying to get at (beware, though – every time you think you’ve “got” the CLT it has a way of revealing an additional layer of intrigue).
This past summer, as I’ve mentioned before, I was teaching school-going students courses in economics, statistics and public policy. I have made use of Seeing Theory in the past, but with the advent of ChatGPT (and especially ChatGPT4), I figured it might be a good time to not just show “cool” visualizations, but also actually try and build them.
And so we did! As we covered a topic, I would ask my students to “build” a working demo using ChatGPT (or Bard). I would nudge and prompt the students to well, write better prompts, and if necessary, step in and write the prompts myself on occasion. But for the most part, the work was done by the students, and we were able to get simple working demos of some stats concepts out of the door.
The “whoa, this is so cool!” moments were worth it in and of themselves, but it is my ardent hope that the students understood the concepts a little bit better for having seen the visualizations.
A great example is the Monty Hall problem. Run a simple Google search for it, if you haven’t come across it before. In my experience, some students tend to not “get” the explanation the first time around. Until this summer, I would get around this problem by asking them “what if it was a million doors instead?”, or if all else failed, by actually “playing” the game using three cards from a deck of cards.
But this time, we built a demo of the problem! So also for Chebyshev’s inequality, the expected value upon rolling a pair of dice and a simple way to visualize what regression does. The demos won’t satisfy professors of statistics or professional coders, for you could add so much more – but for young students who were trying to internalize the key concepts in statistics, it was pure magic.
And the meta lesson, of course, was that they should try and do this for everything! Why stop at stats? Build working demos for concepts in math, in physics, in geography. And if you know even a little bit of coding, try and build even better demos – both I and my students were relatively unfamiliar with coding in general, so we stuck with simple HTML.
But with AI’s new coding capabilities, it is clear that teaching (and learning) can become much better than was the case thus far. If you wish to disagree with me about the word “better”, I look forward to the argument, and you may well end up having more than a couple of points. But the classes were certainly more interactive – and at least along that one dimension, they were certainly better.
I hope to do much more of this in the months and years to come, but for the moment, do try out some of these demos, and let me know how they could be made better.
Let’s begin with the second class today. Your outline mentions the topic “Time Management and Opportunity Cost”. Before we begin the class, outline a definition of both terms, and explain their importance to my life. When you focus on the importance of time management to my life, make sure that you remember I am an eighteen year old. Why should I bother with managing my time? I have my entire life ahead of me – time isn’t a constraint, surely? When then do I need to manage it? Once you explain your answer to these questions, proceed with the outline as discussed, as per the format I have asked for.
You might think this (time management and opportunity cost) to be a weird topic for a second class in a course called “Principles of Economics”. You would certainly think it to be unconventional. Not the latter half of the topic – opportunity costs – but the first one. What does time management have to do with economics? Well, think of it this way – if you are an Indian student who has learnt economics, you have almost certainly come across Lionel Robbins’ definition, and have most likely memorized it back then.
Here it is: the science which studies human behaviour as a relationship between ends and scarce means which have alternative uses.
What is more scarce than time? We all have a limited amount of time, and we all have ends to achieve. The ends we would like to achieve in our lives are much more than the time that is afforded to us, and so we must choose which of these ends to pursue, and which to sacrifice. That’s a pretty good, and if you ask me, useful way to understand opportunity costs. The weirdness, or the unconventional choice (to some) of the topic, is a direct consequence of my request to ChatGPT from yesterday’s post:
Optimize for helping me understand the principles and their applicability to various aspects of my life (professional, personal etc), and it is fine if the outline doesn’t match what is usually taught in introductory economics textbooks. Given this, would you like to change your outline in any way? If yes, please show me the new outline. If no, please begin with the first class.
I am seeking to learn economics not to write an examination at the end of the semester in a college or university with this course. I am seeking, instead, to learn economics in order to apply it to various aspects of my life. Of course, as a student enrolled in a university, you may seek to optimize your learning geared towards doing well in a examination at the end of the semester, and that’s fine. Just let ChatGPT know accordingly, that’s all:
A minor tangent before we get back to our course. What if I asked you to change your course outline for these thirty classes to help me write an examination in an Indian undergraduate course? What would the new outline look like? In your answer, show me both outlines together, per lecture. That is, the first line in the description of lecture 1 (out of 30) should be your original lecture, and the second line should be the course tweaked for doing well in an examination, rather than applicability to various aspects of my life. Ditto for the second course, and so on and so forth for all thirty.
I won’t show you the whole output, but simply how the second lecture’s outline has changed:
Here’s a revised outline for the thirty lectures, with the original lectures designed for applicability to various aspects of life (Lecture A) and the modified lectures tailored to help you succeed in an Indian undergraduate economics course (Lecture B): Lecture A: Time Management and Opportunity Cost Lecture B: Opportunity Cost: Definition, Types, and Examples
ChatGPT-4’s Output (in part)
I know which one I prefer, and why. The good news, as a student, is that you can do both! Learn in order to score well in an examination, and also learn in order to figure out how to apply economics better in the case of your own life. Why should the two be different? Ah, some questions you should reflect upon, rather than ask ChatGPT.
Anyway, back to our lecture du jour. I asked ChatGPT to explain why it chose time management, and I do not think I would have asked that question as an eighteen year old. The older you get, the more aware you are of how limited your time is. And at least in my own case, the converse is also true. I count this as a mark in my favor – that while a good prompt may get a student going, said student will still need help and advice on an ongoing basis.
So far, at any rate.
Further proof of that fact that I’m not out of a job, just yet, is below. The context is that I read the answer, and felt it to be incomplete. So I prodded it a bit, and then just a little bit more:
I’ve often read the phrase “all costs are opportunity costs”. Please explain what this means. Remember that I know no economics, and as usual, give me one example from the Mahabharata, and one from a real-life situation
My professor wondered why the word “TINSTAAFL” hasn’t come up in your answers yet. I didn’t even know this was a word! Is he joking, or is this word relevant to what you’re telling me right now?
To be clear, it isn’t so much about the phrase TINSTAAFL, as it was about the fact that I felt its explanation to be incomplete. This prompted me (no pun intended) to ask it to be more thorough:
I have an ongoing request. I’m looking to make my professors job as easy as possible, or even make him, in some sense, redundant. Optimize your answer for thoroughness, and if you think you can’t fit all of what you want to say in a single output, end with a line that says, “I can tell you more, please let me know if you’re interested to know even more” This will always be applicable in our conversations. Now, back to the second lecture’s outline. Expand upon the three sub-points from the broad contours. I am particularly interested in the third one, so give more details, explanations and background in the case of “recognizing and evaluating opportunity costs in decision making”
This is an important lesson in and of itself. Feel free to tell ChatGPT to give more (or less) detail, or ask it to modify how it gives you the answer (more examples | simpler language | write like person X | show your output as a debate between person X and person Y). Get your “teacher” to be the kind of teacher that you like to learn from!
With regard to your explanation of “Recognizing and evaluating opportunity costs in decision-making”, I’m confused about how to think about short term and long term factors while making my choices, and the short term and long term consequences of my choices. How should I think about this, what framework should I use, and is there an underlying principle at work here that I should know about?
I count this as a pretty important miss on ChatGPT’s part. My personal opinion is that you haven’t fully explained opportunity costs without talking about the importance of how your evaluation of opportunity costs changes given different time horizons. Time matters! ChatGPT actually agrees with me (see below), but only after prodding. And this after making explicit the fact that I was interested in learning about time horizons! And so I asked it again:
Is it useful to think of time preference as a separate principle of economics? More broadly speaking, how should a student of economics think about time preferences? Give me answers from a theoretical perspective, but also from an application perspective.
I’m two days in, where I’m the “student” and ChatGPT my teacher. Today’s class wasn’t great. I don’t think ChatGPT’s output was good enough to stand on its own, and it needed additional prompts to deliver what I would consider to be a good introduction to the concept of opportunity costs, its many nuances and its many applications. It wasn’t bad, but it was far from being good, in my opinion.
Should I take this as a sign that I need to get better at writing prompts, or should I take this as a sign that AI isn’t good enough to replace me yet? How should I change my mental model about whether the average student in a typical college can learn better from AI?
If you are a regular reader of EFE, you know what’s coming next: the truth always lies somewhere in the middle.I need to get better at writing prompts, yes, but also AI isn’t good enough to replace me yet. Both of these things will change over time, of course, but for the moment, less than ten percent into the course, I am inclined to think that I am not out of a job, just yet.
And even better, the complements over substitutes argument just got stronger – I’ll be a much better teacher of a course such as this the next time I get to teach it. Tomorrow we tackle “Supply and Demand: Basics and Market Equilibrium”.
Let’s say you’re a student who is going to start learning economics in the coming semester (starting July 2023). Let’s assume that you’ve never learnt economics in a classroom before, save for a brief introduction to it in high school. If you chose to learn from an LLM instead, how should you go about it?
Leave aside for the moment the question of whether you should be doing so or not. The question I seek to answer over many blog posts is whether you can do so or not. Whether or not this is a good idea for you depends in part on my abilities to add to the value that an LLM generates for you from such a course. And once these thirty (yes, thirty) blog posts are written out, I’ll write about my thoughts about whether a student still needs me in a classroom or not.
My current thinking is that I would still be needed. How much of this is hope, and how much dispassionate analysis is difficult to say right now. For that reason, I would like to tackle this problem at the end of this exercise. For the moment, I want to focus on helping you learn economics by teaching you how to learn it yourself, without the need for a human teacher (online or offline).
In each post, I’ll give you a series of prompts for that particular class. I will not always give you the output of these prompts – feel free to run them as they are, word for word, or tweak them as per your likes, fancies and hobbies.
My motivation in this series is twofold. One, to find out for myself just how much better ChatGPT is than me at teaching you principles of economics. Second, to help all of you realize that you ought to hold all your professors (myself included!) to a higher standard in the coming year. We have to do a better job than AI alone can, along all dimensions – let’s find out if we can.
Buckle up, here we go.
Here’s my first prompt:
Remember, LLM’s work best when you give really detailed prompts. Note the following:
I began by giving some information about myself – my limitations as regards economics, where in the world I come from, and what my interests/hobbies/passions are.
I specified what I’m looking to learn from the LLM.
I specified the quantum of output required (thirty classes).
I specified how broad the output should be.
I specified how I would like the answer to be customized for me
I would like to learn about economics by relating it to what I like to read about in any case (use examples from the Mahabharata)
I would like to learn about economics by relating it to real life situations.
It is amazing to me, regardless of how many times I experience it, that it “gets” what I really mean in spite of having phrased my question using really bad grammar.
The specific examples aren’t the point, the idea is the point. Learn calculus by relating it to mandala art, for example. Learn history by relating it to dance forms. Learn geography by relating it to food from different parts of the world. A teacher in a classroom cannot possibly do this for all the students, because of the size of the class, and because a teacher cannot possibly know your hobby in as much detail as you can. Make good use of AI!
Should the examples from the Mahabharata be chosen for how prominent the examples were in the text, or should they be chosen for their relevance to economics? My preference is for the latter, and I made sure the LLM knows this. Ditto for the real life examples.
I ended with a meta-prompt, that will stay true for the next thirty (or more questions) – ask if I need to learn more, and only then proceed with the next class.
Should you copy this prompt, word for word? Of course not! For one, you may not want to learn economics, but rather a different subject. The underlying principles still holds. You may not like to read about the Mahabharata, for another. You may want only ten lectures, not thirty. Or you may want two hundred! Feel free to tweak the prompt to suit your requirements, but it helps to “get” how to go about thinking about the structure of the prompts. That’s the point.
I took a look at the outline of the thirty course lecture series it prepared for me, and it was not bad at all. But I had a follow-up request:
Now, you might think that you need to know economics in order to judge the output, and tweak your request. And sure, you’re right that it will help. But regardless, even if you cannot judge the quality of the output, surely you know enough about what and how you want to learn. My apologies for going all meta on you, but if you don’t know enough about the supply side of the market, surely you know what you would like as a consumer – at least in part. So feel free to help the LLM become a better teacher by telling it more about you.
It went ahead and gave me the refined output, and also the broad contours of the first class. Here are the broad contours of the first class:
Again, note that I am quite excited about how this class is shaping up, because if economics is, indeed, the study of how to get the most out of life, Arjuna’s choice to fight in the Kurukshetra war is an awesome way to get some really thought-provoking questions in for discussion. But this may not be your cup of tea – so feel free to brew your own cuppa of econ, by customizing it to what you like the most (Avengers? Cricket? RRR? Bharatnatyam? Junk food? Anime? Go for it!)
I did have follow-up questions:
And based upon its answer to this prompt, I had yet another clarificatory question:
Note that your conversation will be (I would go so far as to say should be) different. You will have different questions, different prompts, different things that make you curious. And that’s not just fine, that is the whole point. Depending on how carefully you read its output, and depending on how probing and detailed your questions are, you can keep just this first class going for a long, long time. How long? That’s up to you!
Here are two examples:
You can, of course, ask it to answer any (or all) of these five questions. Ask it to create ten (or twenty, or a hundred) instead – and as a student, assume that this is how us professors might well be “coming up” with questions for your tests, assignments and exams.
Here are more, and note how they get wilder (more random?) with each passing question:
In each of these cases, you don’t have to have trust in, or agree with, the answer given by the LLM. Treat the output as a way to get you to think more deeply, to challenge what has been said, to verify that the answers are correct, and to have further discussions with your peers and with your (human) teachers, whoever they may be.
Note to myself (and to other teachers of an introductory course about the principles of economics):
How can we do a better job than this in the classroom…
Without using AI (we’re substitutes)?
By using AI (we’re complements)?
What is missing from the LLM’s output (this is assuming you’ve tried these prompts or their variants)?
What stops us from recommending that students do this in class on their own devices, and we observe, nudge and discuss some of the more interesting output with everybody? That is, how does teaching change in the coming semester?
Feedback is always welcome, but in the case of the next thirty posts, I think it is especially important. So please, do let me know what you think!
I had (and have) sympathy for Navin, but I have to confess that I did enjoy reading this tweet, because it is very much a teachable moment:
I don't know why I naively expected that if I click on the "Unsubscribe" link on the spam emails @BajajAllianz is sending me, they would actually unsubscribe me.
Because firms have an incentive to make it as difficult for you to “leave”. They make it as easy, painless and frictionless as possible for you to “join”, and they make it as difficult, painful and, well, friction-full as possible for you to leave.
Here’s Richard Thaler and Cass Sunstein on this phenomena:
Perhaps the most basic principle of good choice architecture is our mantra: make it easy. If you want to encourage some behaviour, figure out why people aren’t doing it already, and eliminate the barriers at a standing in their way. If you want people to obtain a driver’s licence or get vaccinated, make it simple for them, above all by increasing convenience. Of course this principle has an obvious corollary: if you want to discourage some behaviour, make it harder by creating barriers. If you want to make it harder for people to vote, forbid voting by mail and early voting, and reduce the number of polling stations (and place them far away from public transportation stops). While you’re at it, try to make people spend hours in line before they can vote. If you don’t want people to immigrate to your country, make them fill out a lot of forms and wait for months for good news in the mail (not by email), and punish them for answering even a single question incorrectly. If you want to discourage poor people from getting economic benefits, require them to navigate a baffling website and to answer a large number of questions (including some that few people can easily understand).
Nudge, by Richard Thaler and Cass Sunstein, Chapter 8, pp 151, Kindle Edition
And they have a term for it too – sludge:
Any aspect of choice architecture consisting of friction that makes it harder for people to obtain an outcome that will make them better off (by their own lights).
Does not getting spam mails in his inbox make Navin better off, by his own lights?
Yes, of course!
Does the design of the unsubscribe (I’m being generous here) form add friction to the process of Navin obtaining this outcome?
Yes, of course!
That’s sludge in action.
And once you “see” it, you begin to spot it everywhere. Newspapers and magazines make it difficult for you to cancel your online subscriptions and banks make it difficult for you to file a complaint with the banking ombudsman, to give you just two examples. I’m sure you can think of many more from your own life, and Chapter 8 of the book Nudge has many, many other examples. Please read the whole chapter (and if you’re willing to humor me, the whole book).
And finally, this might resonate with people of a certain age (or maybe, even now, all ages?):
If you have the Monday blues, and now have an irresistible urge to drop everything else and watch the whole episode instead, it’s S04E04.
“The 1960 vs. 2022 nature of graph and the 1-6 ratings of income are quite confusing”
“The “middle income trap” is too dense and you pointing to Nicaragua shows that the journeys of different countries could be going in different directions within that group”
“The range of 1.75-3.75 on both axes is deceptive though. While the higher scales 4+ is rich in general, the relative gap between India/Nigeria and China is very high — not sufficient distinction there.”
“Putting China in “middle income trap” is odd because it has gone from below 2 in 1960 to above 3 in 2022 (based on the axes).”
Before I get around to answering his questions, I have a question for you.
Do you have any questions of your own, for having read his questions? Go read my post again, stare at the chart, go over Mihir’s questions, and then think about whether you have any questions of your own.
I’ll answer each of Mihir’s questions below, but the point of this post is really what follows after, so please do stick around until the end!
“The 1960 vs. 2022 nature of graph and the 1-6 ratings of income are quite confusing”
Yup, absolutely. It takes a while to figure out what is going on in an Economist chart, and while that is a problem, I’d argue that the rewards are usually worth it. By the way, if you are an Economist subscriber, you absolutely should read their newsletter on visualization and charts.
A useful principle to keep in mind is that when you look at a chart, train yourself to not look at the data first. First be clear about what is on the axes (all of them). Then be clear about the title of the chart. It helps to take a look at the source of the data. Then start taking a look at the chart itself.
Homework: what does “income per person, relative to the United States, log of %” mean? Can you explain this phrase to somebody else? If you can’t, you haven’t understood it well enough!
“The “middle income trap” is too dense and you pointing to Nicaragua shows that the journeys of different countries could be going in different directions within that group”
The central square in the chart is too dense, but that’s just fine by me. Why? Because the outliers are then even more worthy of analysis. If you cannot “make it” into the central square, then you’re even more special relative to that crowded space. Botswana is special because it was poor in 1960, and is not just middle-income today, but on the verge of breaking into the high-income space. That’s a special story!
Argentina, on the other hand, is special for the wrong reason. It was a high-income country back in 1960, but has since slid down into a middle-income country grouping. Both of these countries, within the context of this chart, also help you understand Mihir’s second comment here. Because this is a static image, and because we’re comparing two different points in time, we don’t get a sense of the trajectory of a country. Botswana is on the way up, and Argentina has slid down – but you need to know this separately. This isn’t clear from looking at the chart.
To be clear, this isn’t a criticism of the chart, but rather a way of recognizing that your work as a student doesn’t stop for having studied the chart. Au contraire, this chart should spur you to read more about whichever country seems interesting to you.
“Tell me more about Botswana’s growth story over the last sixty years or so. Assume I know very little about Africa in general, and Botswana in particular. Your answer should include Botswana’s internal politics, key leaders, relationship with her neighbors and with the superpowers during the cold war, her natural resources and some background on major ethnic and religious groups in Botswana”
“The range of 1.75-3.75 on both axes is deceptive though. While the higher scales 4+ is rich in general, the relative gap between India/Nigeria and China is very high — not sufficient distinction there.” Log scales can be tricky, and the best way to understand this is by thinking about how earthquakes are measured. And yes, Mihir is spot on about how you need to keep this in mind. The lower ends of the middle income square (left to right and bottom to top) actually cover very large ground, and countries in the left-bottom corner are very different from countries in the right-top corner of the middle square. Dividing the middle square into a 3×3 grid would be a great idea. (Hi, The Economist. Hint, hint)
“Putting China in “middle income trap” is odd because it has gone from below 2 in 1960 to above 3 in 2022 (based on the axes).”
It’s their chart to make, and ours to interpret as we see fit, so while I get where Mihir is coming from, I’m fine with both the boundaries of the middle square, and with the framing that The Economist has used. China’s growth trajectory over these past sixty years or so has been fantastic, but the question is about whether it can keep that break-neck growth rate up going ahead. A very wise economist won a Nobel Prize for coming up with a simple model that says “Nah, probably not”. So while I understand Mihir’s point, I can see the logic used by The Economist as well. Stop me if you’ve heard this before, but macro is hard.
But now that I’ve replied to his comments, let me come to the main point of today’s post.
What stops us from asking questions as we read? Why, that is, do we read unquestioningly?
Maybe that’s too specific a question, so let me step back and frame it more generally.
Why don’t you argue more often with whatever you’re reading?
Make sure you understand where they’re coming from, and that you understand their line of reasoning, to the extent possible. And also that you understand how and why they reached the conclusions they did. You don’t have to agree with either the line of reasoning or the conclusion, to be clear.
But asking smart, probing questions about both the premises and the conclusions can help you become a much more engaged reader. This, in turn, can help you to both understand what you’re reading, and to decide whether you agree with the author.
It’s a rare old skill, and I’d encourage you to apply it, always, while you’re reading.
So please, disagree more with what you read on my blog, and let me know of your disagreements.
But first things first: what is a middle income trap?
Here’s my assistant’s answer:
The middle-income trap is a concept that describes a situation where a country’s economy gets stuck in a middle-income range, and it struggles to advance to higher levels of development. To understand this, we need to first know about development economics and growth theory. Development economics is the study of how countries become richer and more developed over time. Growth theory is a part of development economics that studies how economies grow and what factors help or hinder growth. Now, let’s go back to the middle-income trap. When a country starts to develop, it usually grows quickly and becomes a middle-income country. However, some countries find it hard to keep growing and end up getting stuck in this middle-income range. The problem is that these countries face new challenges when they reach this stage. For example, they may have to compete with cheaper goods from other countries, or they may need more advanced technology to keep growing. These challenges can be difficult to overcome, and the country’s growth slows down or even stops. This can be a problem because being stuck in the middle-income trap can mean that a country’s people do not get richer over time, and the country may not be able to provide the same opportunities as wealthier countries. To avoid the middle-income trap, countries need to keep innovating and adapting to new challenges. This can involve investing in education and training, creating new technologies, and finding new ways to do things. By doing this, a country can keep growing and move towards becoming a high-income country.
So which countries find themselves in the middle income trap? The Economist has a lovely chart within an article that helps us find out the answer to this question
As a country, you want to be in the top row, since that implies you are doing well today. You do not want to be in the bottom row, since that implies that you are not doing well today. Things get interesting in the row in the middle.
The first column (read left to right) in the middle row is a great story. You were “low” in 1960, and are “middle” today, so you’ve covered impressive ground, and are hopefully on your way up from here on in. The third column (read left to right) in the middle row, is a problematic story (to put it mildly). You were “high” in 1960, but have reached “middle” today.
The square bang in the middle? That’s the middle income trap.
These are countries that were doing relatively well back then, in 1960, and are doing relatively well today – but in the sense that there hasn’t been a relative improvement, they find themselves in a middle income trap.
What does “relatively well” mean, and what does “relative improvement” mean? Take China as an example – China was a middle income country in 1960 (it falls in the middle, read left to right), and it is a middle income country today (it falls in the middle, read bottom to top). That’s not to say that there has been no improvement for the Chinese since 1960, of course! There has been remarkable improvement.
But relative to the USA, China was in the middle of the pack in 1960, and finds itself to be middle of the pack in 2022. Therefore the middle income trap.
Even within that square itself, by the way, there are stories to be discovered. If you are in the right top of that square (Mexico, for example), you were on the verge of becoming a high-income nation in 1960, and you are on the verge of becoming a high income nation today, but you haven’t actually achieved that status as of yet.
If you are in the top left of that square in the middle, you were barely better than low income in 1960, and are about to break through the metaphorical ceiling today (China, for example). India hasn’t moved much within that middle square, and that is therefore a frustrating story for us in India.
Homework: how would you describe Nicaragua’s position in this graph? Is it better off or worse off over these past sixty years or so?
Finally, you might also want to think about whether the middle income trap is such a bad thing in the first place!
Poland and Malaysia may now be running into this [he’s referring to the middle income trap here – Ashish] problem. McKinsey cites Poland’s need to develop or acquire strong brands in order to catch up with West Europe. The failure of Malaysia’s attempt to build domestic champions is worrying. And yet I see two responses to this. The first is: Do we really care? Poland and Malaysia may not be as rich as Germany or Korea, but they’ve definitely escaped poverty. Countries like Bangladesh or Vietnam or Ghana or even Mexico would kill to have a per capita GDP of $30,000. That’s about the GDP of the U.S. in the early 1980s. Is it really fair to call that level of development a “middle income trap”? If you’re a poor country, and you have a reliable, dependable way of getting as rich as the U.S. was in the early 1980s, dammit, you take it. You don’t worry about whether that strategy will eventually make it harder to get as rich as the U.S. of 2023.
Let me be clear – I am not saying (and I don’t think Noah is either) that more growth is a bad thing. But a targeting of rapid growth at all costs, and above all else, isn’t necessarily a great idea.
What is the optimal amount of a problem – any problem – in your life? Surely the answer to this question must be zero! No?
Let’s get more specific. What is the optimal amount of poison in your body? Surely the answer to this question must be zero! No?
Well, that would mean no vaccinations, for starters. And no sugar in your body, but that’s a whole other story.
But it certainly would mean no vaccines in your body. Here’s the definition of a vaccine:
“A substance used to stimulate immunity to a particular infectious disease or pathogen, typically prepared from an inactivated or weakened form of the causative agent or from its constituents or products.”
So if you want your body to develop the ability to fight a particular virus, you first need to inject your body with a (much) milder form of that virus. The optimal level of “poison” in your body? Non-zero!
But why should the optimal level of fraud be non-zero?
The answer, as it turns out, is related to opportunity costs:
All fraud is a) an abuse of trust causing b) monetary losses for the defrauded and c) monetary gain for the fraudster. You could zero fraud by never trusting anyone in any circumstance.
Is it possible to have a system set up where there is no fraud whatsoever? Sure. You’re just likely to have a bunch of extremely angry customers, because you’ll have to set up systems where they have to verify that they are who they say they are every few seconds.
Logged in from the same device? I don’t care, you must sign in again. Transacting for a slightly larger amount than usual? I don’t care, you must verify yourself three different ways. Reset your password every three days, and if you don’t comply, we won’t let you in.
Turns out you can live in a zero-fraud world within your organization. It’s just that nobody will want to buy anything from you:
Clearly, e-commerce would cease if, prior to buying a pair of sneakers online, you required someone to go to that degree of effort. You’d almost never lose a pair of sneakers to a fraudster again, but you’d also sell very few sneakers.
And so, much like vaccines and the human body, organizations often voluntarily accept some fraud. Why? Because of what we spoke about above – eliminating all fraud risks throwing the baby out with the bathwater. That is, yes, you could eliminate all fraud, but you would also, in the process, end up turning away all of your genuine customers as well. Or, if you prefer it to be put in our language, the opportunity cost of zero fraud is zero customers.
Which then begs the question – how much fraud is acceptable? Well, that would depend on the volume of transactions, and on your margins. High margin, high volume business? Fraud isn’t your first problem. Low margin, low volume business? Fraud probably makes it into your top three list of problems.
Between these two there exists a spectrum of fraud regimes, and this is broadly a good thing. Society gets to make choices, and here it is choosing through the activities of private agents. It is optimizing for how many resources to let leak to bad actors and much societal effort to burn on policing them versus how much low-friction commerce to enable by good actors. This is often missed in discussions of fraud; one reason it has increased over the past few decades is that legitimate commerce has exploded, as the world becomes richer and as barriers to commerce have come down.
1/12 I fully agree with Jim O'Neill that "the dollar plays far too dominant a role in global finance", and that it would be better (for the US above all) if the BRICS were able to manage to cut back its role dramatically. https://t.co/xqgHdbVUgP
Therefore, we call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4. This pause should be public and verifiable, and include all key actors. If such a pause cannot be enacted quickly, governments should step in and institute a moratorium
Lots of different ways to think about this, and as always, the truth lies somewhere in the middle. But forget all of the arguments for the moment, homilies from a variety of different languages go a long way towards helping you understand that this is a (mostly) lost cause. You can talk of hungry sparrows in the field in Hindi, or you could talk about getting the genie back in the lamp in English. You might as well talk about saying “statue!” to a tsunami, and you might actually have better luck with that plan.
But LLM’s are here, they’re about get better capabilities, and they will be used for good and for bad.
That’s it. C’est tout.
As with everything else, there is a lot to read about this issue, but there are two pieces in particular that I enjoyed reading. The first is a piece written by Sayash Kapoor and Arvind Narayanan in their excellent newsletter, AI Snake Oil. Worth subscribing to, if you ask me.
The letter positions AI risk as analogous to nuclear risk or the risk from human cloning. It advocates for pausing AI tools because other catastrophic technologies have been paused before. But a containment approach is unlikely to be effective for AI. LLMs are orders of magnitude cheaper to build than nuclear weapons or cloning — and the cost is rapidly dropping. And the technical know-how to build LLMs is already widespread.
LLM’s are already in the wild, they can now run on devices manufactured three years ago, and the models will likely become more efficient over time, and hardware capabilities will get better over time. It’s all well and good to want to pause, but I don’t think the letter spends nearly enough time in asking “how”, let alone answering the question.
Speaking of omissions from the letter:
Is there any mention of public choice/political economy questions in the petition, or even a peripheral awareness of them? Any dealing with national security issues and America’s responsibility to stay ahead of potentially hostile foreign powers? And what about the old DC saying, running something like “in politics there is nothing so permanent as the temporary”? Might we end up with a regulatory institution as good as the CDC?
You know you’re in trouble when Tyler Cowen decides you’re worthy of some gentle trolling.
But on a more serious note, the meta-lesson here is that if you are going to recommend a particular policy, you’d do well to ask how feasible it is in the first place. There is always the temptation to imagine the end-state Utopia when you make a recommendation. Fixating on that Utopia often distracts us from asking which route to take to reach said Utopia. And every now and then, one realizes that there isn’t any route available at all.
Outcomes over intentions!
One final point: I mentioned that the trust lies somewhere in the middle. In the context of this post, what does this mean, exactly? Should we stop or not? Well, as I’ve explained, I don’t think we can stop – but there is merit to the idea of proceeding cautiously.
